Exploring Machine Learning Approaches Speech Emotion Recognition System

Exploring Machine Learning Approaches Speech Emotion Recognition System

G.Yaswanth,D.Jaya Manjunath,S.Gopi,P.Priyanka

Dept.Artificial Intelligence & Machine Learning

Sasi Institute of Technology & Engineering

Tadepalligudem,India

Abstract— The basis for human-machine interaction depends on speech recognition technology that utilizes two platforms between virtual assistance systems and voice command interfaces. The performance of a machine learningbased speech recognition system is improved by integrating MFCC with both Mel Spectrogram and Chroma features as per research findings. The study examines the individual information extraction capabilities of MFCC and Mel Spectrogram and Chroma features regarding spectral and pitch-related speech signal characteristics. The features provide input to several machine learning algorithms which include k- nearest neighbors (KNN) and support vector machines (SVM) and Random Forest and multi-layer perceptron (MLP) and Naive Bayes (NB). The research conducts evaluation testing to discover the optimal combination of features plus their collective effects on speech recognition accuracy and stability. The study extends its research by examining how emotions can be detected from spoken communication signals. The research examines seven distinct emotion categories including angry as well as disgust alongside fear together with happy alongside neutral followed by surprise and sad. Flask develops a user-friendly interface that enables real-time emotion prediction with the most successful machine learning model discovered for this application.

Key components: Speech recognition, Audio feature extraction and the implementation of Mel Frequency Cepstral Coefficients (MFCC) and Mel Spectrogram and Chroma features.